Uno: A One-Stop Solution for Inter- and Intra-Datacenter Congestion Control and Reliable Connectivity

TL;DR

Uno addresses the challenge of co-managing intra- and inter-datacenter traffic by unifying congestion control and reliability into two components: Uno CC for rapid, fair, ECN-based control, and Uno RC for loss-resilient, multi-path routing with erasure coding. The system introduces phantom queues to emulate large inter-DC BDPs, enables Quick Adapt for extreme congestion, and combines block-level erasure coding with UnoLB load balancing to tolerate failures and improve routing. Across simulations, Uno yields lower mean and tail latencies and faster convergence to fair bandwidth shares compared to state-of-the-art baselines like Gemini and MPRDMA+BBR, and shows strong resilience under failure scenarios. This unified approach offers practical benefits for coexistence of intra- and inter-DC traffic and has positive implications for future cloud and AI workloads that span multiple datacenters.

Abstract

Cloud computing and AI workloads are driving unprecedented demand for efficient communication within and across datacenters. However, the coexistence of intra- and inter-datacenter traffic within datacenters plus the disparity between the RTTs of intra- and inter-datacenter networks complicates congestion management and traffic routing. Particularly, faster congestion responses of intra-datacenter traffic causes rate unfairness when competing with slower inter-datacenter flows. Additionally, inter-datacenter messages suffer from slow loss recovery and, thus, require reliability. Existing solutions overlook these challenges and handle inter- and intra-datacenter congestion with separate control loops or at different granularities. We propose Uno, a unified system for both inter- and intra-DC environments that integrates a transport protocol for rapid congestion reaction and fair rate control with a load balancing scheme that combines erasure coding and adaptive routing. Our findings show that Uno significantly improves the completion times of both inter- and intra-DC flows compared to state-of-the-art methods such as Gemini.

Figures (13)

• Figure 1: A shows inter-DC links for Azure in Europe and their delay assuming point-to-point connections. B shows that inter-DC links make even medium-large messages latency-bound.
• Figure 2: To ensure efficient communication within and across datacenters, Uno integrates congestion control, load balancing, and loss resiliency.
• Figure 3: Gemini and MPRDMA+BBR fall short in efficiently achieving bandwidth fairness while Uno provides fast convergence to fairness during a mixed incast scenario.
• Figure 4: Showcasing the effect of phantom queues for intra-DC traffic. A shows the queue at the incast receiver over time without phantom queues and with phantom queue ( B). C shows the FCTs for the intra-DC flows.
• Figure 5: Uno's overall architecture.